Handle relaunch after application crash?

[caulfield2]

I recently ran into a situation where my Windows application crashed during a test, causing all subsequent tests to fail. Has anyone designed some test logic to check if the application is active at the end of each test and relaunch if appropriate before the next test starts? I'm working in C# using the Visual Studio Unit Test project format.

[PandaMagnus]

Could you check to see if the application process exists during test cleanup and then launch as needed? Basically this, but use WAP to re-launch:
https://stackoverflow.com/questions/4453805/c-sharp-check-if-a-process-exists-then-close-it

It might be better to do it on test setup, actually. That way if it fails on the last test, you're not relaunching the application before tearing down your whole test run.

[caulfield2]

@PandaMagnus Definitely better to do before each test. I like the idea of checking running processes, but I was hoping to find a way to check using the Selenium/Appium libraries. This is because I test my tests on a remote machine when I am writing them and Process.GetProcessesByName() would return the processes on my machine and not my test machine. This would work with a CI pipeline, however, as the tests are executed locally on the test machine.

[PandaMagnus]

@caulfield2 aaah, okay understood. Could you invoke a known property on the driver, either in a try/catch or a conditional check? I'm not as familiar with Appium, but in Selenium if I invoke, say, Driver.WindowHandles or Driver.PageSource after I kill a browser, I'll get a return of 0 or NoSuchWindow exception (respective). So, E.G., in test setup:

if(Driver.WindowHandles == 0)
{
   // Re-launch the application, maybe re-instantiate a new driver, etc.
}

or...

try
{
   var _ = Driver.Pagesource;
}
catch(NoSuchWindowException)
{
   // Re-launch the application, maybe re-instantiate a new driver, etc.
}

Unfortunately I don't have WAP at work yet, so I can't check myself. :( If I get to it tonight or tomorrow, I'll check and get back to you.

[Manvibha]

I would suggest not develop such dependency between test scripts; dependency would lead to such expected issues. Rather, make each scripts independent of each so that if application fails/crashes due some reason, next script will launch the app and continue. I know this would increase the test suite execution time but this would ensures that all scripts are executed for sure.
I follow a rule that start and end point for each script should be same like Script starts with launching APP and login , then close at the end of each script finishes (Test clean up).

Mahendra